What's up with multi core really poor performance under 128 buffer?

[TNM]

edit, not worth feeding the troll. I have learned a very valuable lesson this past week here.. From now on, i am not going to give confrontation seekers any of my time. I'll insta block, they can say the nastiest things in the world about me all day every day for all i care, and i will never see a word of it. If i don't respond no matter what they say, it will not only drive the trolls insane, but it will mean i can not be banned whatsoever, as I have not engaged them.. starting right now

[TNM]

Originally Posted by bartosz idzi

Hi, i attached my results. I'm on Dual 2,26 5520 Xeon 2009 Mac Pro with RME Hammerfall card under OS X 10.9.5. PT 12.7.1 @32 samples buffer.

Fantastic news.

So i presume the 3ghz mac pro I am looking at from 2013 (it's either that or windows, i still haven't decided), won't have the issue.

Firstly, THANK YOU for staying on topic and doing this. Very kind of you.

If you ever need a favour please ask any time, anything I can do to help.

RME have exceptionally good drivers, so this goes to show, there are a few variables at play here

1) Sierra is the issue
2) PT 12.8 changed something
3) The macbooks included on board sound/pro tools aggregate driver has an issue
4) The macbook's processor frequency cycling is the issue

It's quite possible that even if i get a new beast computer, or even an imac pro, that with sierra/high sierra, and PT 12.8.3, the issue will persist.

i will see if i can get to an avid dealer that has a modern system set yo and test it in person..

May I ask, what VI you had armed when doing this? I'd like to try the same VI here at 32 buffer to take a direct comparison.

Also, just to confirm, you were playing midi notes yes?

Cheers!

Edit, i don't think this is about upgrading my mac after all, i think it's an OSX issue.. I will know myself in a few days as my imac is able to take mavericks.
So i will install mavericks and my apollos, and PT, and do a test. If fixed, i will know for a fact that sierra and 12.8 changed something, therefore i can report it to avid as a bug.
i'll update this topic when i have the answer.

Edit 2: one more interesting thing, according to benchmarks, my laptop cpu or even my imac 2600K, is roughly equivalent to 2 of your xeons, so since you have 2, we have approximately the same power, with the difference being that my single core performance is a lot stronger (and it's single core where this should traditionally matter, in monitoring a single instrument, as 95% of VI's can not access multiple cores nor can/should DAW do so). I bet you if you ever upgrade to sierra and 12.8 the problem will happen for you, so my advise is, don't, for now at least!

If i get a new mac I won't have a choice sadly. Hence why windows is becoming an interesting option a custom built 7820X beast.

[TNM]

Originally Posted by Drew Mazurek

PT has never been anywhere near as CPU efficient as other DAWs. There are many competing theories as to why, but if this is news to you, you haven't done your research.

i just loaded up my yosemite and 12.4 partition on my imac, from a usb3 flash drive i saved it on (64 gb sandisk ultra), and the issue is not there.

Yosemite and PT 12 First is what i tested Pro tools demo on 2 years and a week ago when i first bought Pro tools, and i got familiar with first before i bought. So i DID do my research thanks, and would have remembered such an issue.

Once again, instead of going off topic, this is not about pro tools overall performance, which is fine here, but why not test to see if your machine engages multiple cores when trying to play just one vi at low buffer? Stay on topic and see if it happens or not.. This is not about PT's overall performance, this thread is about one specific issue which I believe is a coding bug that made it into PT 12.8, where multiple cores are engaged for a task when they shouldn't be.

[Drew Mazurek]

Originally Posted by TNM

this is not about pro tools overall performance

It is for me. I'd like it to be better. At least on par with other DAWs.

[grayter1]

Originally Posted by TNM

This is not about PT's overall performance, this thread is about one specific issue which I believe is a coding bug that made it into PT 12.8, where multiple cores are engaged for a task when they shouldn't be.

It's all very hit and miss. Kontakt, Maschine and Sampletank will sometimes tax the cores hard, and sometimes they don't. Sometimes an instance of Kontakt will hit really hard, then fall back to low levels. It all depends on the programming (and internal effects) around the patch. Also having the instrument track selected can make a huge difference in core usage, and simply unselecting an instrument track will cause the CPU to go down.

There really is no rhyme or reason to the CPU meters in PT, and I've stopped using them, so that was good advice given by the others here.

I too use Logic for composition. I've been using it since 1998, and again as stated, it's ALWAYS been better at hosting VI's. Reaper is fantastic as well.

Yes, it's probably a bug in Tools, but it's been that way for so long, and I don't expect it to change any time soon. That said, my Hackintosh (see specs below) performs better than my old iMac, and 12.8.3 (for ME) has been the best version yet for virtual instruments. I've been working for the last 2 weeks (non-stop) with VI's at 64 samples with no errors or hiccups.

I believe that's because my system is up to the task, along with some improvements by Avid. Of course, YMMV.

tg

[john1192]

Originally Posted by grayter1

It's all very hit and miss. Kontakt, Maschine and Sampletank will sometimes tax the cores hard, and sometimes they don't. Sometimes an instance of Kontakt will hit really hard, then fall back to low levels. It all depends on the programming (and internal effects) around the patch.

There really is no rhyme or reason to the CPU meters in PT, and I've stopped using them, so that was good advice given by the others here.

I too use Logic for composition. I've been using it since 1998, and again as stated, it's ALWAYS been better at hosting VI's. Reaper is fantastic as well.

Yes, it's probably a bug in Tools, but it's been that way for so long, and I don't expect it to change any time soon. That said, my Hackintosh (see specs below) performs better than my old iMac, and 12.8.3 (for ME) has been the best version yet for virtual instruments. I've been working for the last 2 weeks (non-stop) with VI's at 64 samples with no errors or hiccups.

I believe that's because my system is up to the task, along with some improvements by Avid. Of course, YMMV.

tg

good to hear about VI performance .. real world tests are the only way to know this ..

and BTW: time to update ypour signature to 12.8.3 !!!

[grayter1]

Originally Posted by john1192

good to hear about VI performance .. real world tests are the only way to know this ..

and BTW: time to update ypour signature to 12.8.3 !!!

I did it as you were posting

[musicman691]

Originally Posted by TNM

i just loaded up my yosemite and 12.4 partition on my imac, from a usb3 flash drive i saved it on (64 gb sandisk ultra), and the issue is not there.

Yosemite and PT 12 First is what i tested Pro tools demo on 2 years and a week ago when i first bought Pro tools, and i got familiar with first before i bought. So i DID do my research thanks, and would have remembered such an issue.

Once again, instead of going off topic, this is not about pro tools overall performance, which is fine here, but why not test to see if your machine engages multiple cores when trying to play just one vi at low buffer? Stay on topic and see if it happens or not.. This is not about PT's overall performance, this thread is about one specific issue which I believe is a coding bug that made it into PT 12.8, where multiple cores are engaged for a task when they shouldn't be.

This IS about PT performance as that's the program running when you're supposedly finding the issue. Does the same machine do the same thing when using another daw? If not - that's putting the gaze on PT. I seriously doubt this is due to running Sierra.

Like I wrote before - this is why a lot of people use VEPro. That YOU had a problem with it doesn't negate the fact using it has helped a lot of people get good core usage on their computers.

You'll never get single core usage when running even one vi. Typically I find anywhere from 3 to 4 cores being used when running a single vi. Usually one core will run a little more usage than the others but never over 10%.

Besides who really needs to run at 32 samples when running at 128 or above seems to work for you?

[TNM]

Originally Posted by grayter1

It's all very hit and miss. Kontakt, Maschine and Sampletank will sometimes tax the cores hard, and sometimes they don't. Sometimes an instance of Kontakt will hit really hard, then fall back to low levels. It all depends on the programming (and internal effects) around the patch. Also having the instrument track selected can make a huge difference in core usage, and simply unselecting an instrument track will cause the CPU to go down.

There really is no rhyme or reason to the CPU meters in PT, and I've stopped using them, so that was good advice given by the others here.

I too use Logic for composition. I've been using it since 1998, and again as stated, it's ALWAYS been better at hosting VI's. Reaper is fantastic as well.

Yes, it's probably a bug in Tools, but it's been that way for so long, and I don't expect it to change any time soon. That said, my Hackintosh (see specs below) performs better than my old iMac, and 12.8.3 (for ME) has been the best version yet for virtual instruments. I've been working for the last 2 weeks (non-stop) with VI's at 64 samples with no errors or hiccups.

I believe that's because my system is up to the task, along with some improvements by Avid. Of course, YMMV.

tg

well sorry but I just don't experience that here when the buffer is 128 or higher. I have done hardcore DSP tests using identical effects and pro tools always outperforms studio one and Cubase, playing back identical files with identical routings and identical VI's and FX. Basically it's a 32 track dsp tester project i have designed to test all the DAW's, and every track is routed to an aux rather than the master, then some are routed to a few various aux's, there is a smattering of fx carefully designed to tax every core. Only Logic and Reaper had better performance. I have Cubase Asio guard on medium to do the test, and S1 dropout protection on high.

As far as a track being selected, here it's set up that i can select anything at any time, and, unless i manually press the record or monitor button, only then pro tools puts those tracks into low buffer mode. I really presumed it was the same for everyone and that's how I like it.. with logic you can NOT select an instrument track without it going into low latency mode which is really problematic in heavy projects believe me.

I really feel we are getting our wires crossed here, and even the mod is insisting on changing the topic.. sure, we can all wish our DAWs to perform better and optimally at all times, and besides this topic, if there is a flaw in pro tools' engine that I hadn't noticed before, thanks for pointing that out.

But once again, this is about a specific behavioural change in Sierra and 12.8, where as engaging just one instrument track in live buffer mode (i;e so it can be played from a midi KB), is engaging all cores equally for ONE instrument, which actually shouldn't even be possible, hence why I say it's a bug. Check out the only user who actually answered on topic's pic, and you will see that in mavericks and PT 12.7 it doesn't happen.
It's 100% reproducible here when using 12.8 and sierra.

Once again, and I don't know how many times I am going to have to repeat this, istat and activity monitor confirm pro tools meter, and show that all cores are absolutely equally engaged. These are REAL CPU meters that are correlating with pro tools's own meter.. so even if pro tools meter is flaky and not useful generally, in THIS PARTICULAR SCENARIO, it is accurately representing the problem at hand. I have to be completely honest and say that I think most people that have replied so far simply do not even understand what, cause there seems to be some absolute refusal to spend 60 seconds time to see if your machine exhibits the problem, and would rather spend more than 60 seconds making large posts that are entirely unrelated to *this particular issue*. Also, respectfully to the mod, please stop trying to make my topic about something else just cause you want to, why not start your own topic if that's the case?

[karp47]

I'm having a very similar issue. As previously mentioned, it most likely has something to do with PT 12.8.

If I set my buffer to 256 samples then Pro Tools behaves fairly well. As soon as I am on 128 or 64 I get a lot of CPU usage - even if there is only one track in the session. Please note - this is only happening on record enabled tracks.

I have tested this across audio and instrument tracks and have found it to be the same across both.

The only solution I can think of at the current time is to roll back to 12.7.

----

In addition, I have run the same tests in Logic X and have not experienced the same problem. I also do not recall this in previous versions of PT. Leading me to believe that the issue is stemming from PT12.